cff-version: 1.2.0
message: "If you use haptal-curate in your research, please cite it as below."
type: software
title: "haptal-curate: Quality Scoring and Curation for Robot Demonstration Datasets"
version: 0.1.0
date-released: "2026-06-10"
url: "https://github.com/haptal-ai/haptal-curate"
license: MIT
authors:
  - family-names: Bedi
    given-names: Aarav
    email: aarav@haptal.ai
    affiliation: Haptal AI
keywords:
  - robotics
  - imitation learning
  - dataset curation
  - quality metrics
  - demonstration datasets
abstract: >-
  haptal-curate is a Python package for automated quality scoring and curation
  of robot demonstration datasets. It provides a modular suite of quality
  metrics (smoothness, gripper timing, entropy, isolation forest, k-NN,
  trajectory alignment, and a weighted ensemble), phase-gated scoring for
  manipulation tasks, and automatic length-confound detection via Spearman rank
  correlation.
references:
  - type: article
    title: "SPARC: An appropriate smoothness measure for robotic paths"
    authors:
      - family-names: Balasubramanian
        given-names: Sivakumar
    journal: IEEE Robotics and Automation Letters
    year: 2015
  - type: software
    title: "LIBERO: Benchmarking Knowledge Transfer in Lifelong Robot Learning"
    authors:
      - family-names: Liu
        given-names: Bo
    year: 2023
    url: "https://github.com/Lifelong-Robot-Learning/LIBERO"